MODULAR BIO-BEHAVIORAL CLOSED-LOOP CONTROL OF TYPE 1 DIABETES PROJECT SUMMARY / ABSTRACT By nature, the development of a closed-loop system (known as artificial pancreas) controlling blood glucose (BG) in diabetes is an interdisciplinary project involving physiology, behavioral science, and engineering. Consequently, this project represents an interdisciplinary and international effort of physicians, psychologists, mathematicians, and engineers from the United States, Italy, and France dedicated to the understanding of behavioral and biological prerequisites to individually-tailored closed-loop control of type 1 diabetes (T1DM). The principal idea of the proposed research is: in order to be successful, closed-loop control must adapt to individual physiologic characteristics and to the behavioral profile of each person. The keys to this adaptation are biosystem (patient) observation and modular control. Thus, we propose to lay the foundation for a modular system comprised of algorithmic observers of patients' behavior and metabolic state, and control modules responsible for insulin delivery and hypoglycemia prevention. Building this system, we will utilize our extensive expertise with in silico modeling and simulation of the human metabolism, and the experience from our recent closed-loop control studies. The development and testing of this system will be accomplished in four phases: Phase 1 will investigate patterns of behavioral events relevant to T1DM control. A field study will record meals, insulin injections, and exercise in parallel with continuous glucose monitoring, aiming to develop a learning algorithm - behavioral observer - which will track over time key recurrent elements of a person's routine. Phase 2 will investigate relationships of insulin sensitivity and impaired counterregulation with BG variability, aiming to develop algorithmic physiology observers, which will track specific parameters of glucose variability and recurrent hypoglycemia as markers of change in a person's insulin sensitivity and counterregulatory ability. Phase 3 will test in the field an advisory system providing personalized feedback to people with T1DM. The system will be informed by behavioral and physiology observers and will consist of three advisory modules: (i) evaluation of risk for hypoglycemia 24 hours ahead; (ii) bolus calculator suggesting pre-meal insulin doses, and (iii) basal rate advisor suggesting basal rate profiles for the next 24 hours. Phase 4 will focus on automated closed-loop control conducting a series of studies (outpatient & inpatient) testing sequentially three control modules responsible for: (i) detection and prevention of hypoglycemia 1-2 hours ahead, (ii) control of pre-meal insulin boluses, and (iii) control of basal rate and overnight steady state. We envision that, depending on patients' or physicians' choice, each observer or advisory/control module could be used separately, or within integrated open- or closed-loop control systems. A modular approach will permit incremental testing and deployment of system features, which will structure and facilitate system development.